analyze

/analyze — Analyze Post

Analyzes post performance, detects patterns, and sharpens the ICP. Lifecycle-aware: distinguishes Active/Cooling/Archived posts.

IMPORTANT: Delegate the work to the post-analyzer agent. Do NOTHING yourself — start the agent with the Agent tool and pass the post URN.

Usage

/analyze <urn>        # Analyze single post
/analyze              # All posts that need analysis
/analyze patterns     # Pattern refresh only (archived posts)
/analyze icp          # ICP sharpening only

Lifecycle Awareness

Lifecycle	Analysis Behavior
Active (0-7 days)	In data/posts/ — analysis possible, but with caveat "metrics still moving"
Cooling (7-14 days)	In data/posts/ — good data basis, snapshot 2-3 available
Archived (14+ days)	In data/posts/archive/ — final metrics, best basis for pattern detection

Pattern refresh uses only archived posts (data/posts/archive/).

Workflow

Single Post (/analyze )

1. Start post-analyzer agent
2. Result:

Post Analysis: "Most SMBs underestimate..."
URN: urn:li:activity:7435982583777169408
Lifecycle: Cooling (Day 9)

Performance (Snapshot 2):
  Reactions:       89 (Baseline: 45) → +98%
  Comments:        12 (Baseline: 5) → +140%
  Impressions:  3,200 (Baseline: 1,800) → +78%
  Engagement Rate: 3.2% (Baseline: 2.8%) → +14%

Content Properties:
  Hook: Surprising Fact | Format: Text | CTA: Question
  Day: Tuesday | Hour: 8 | Length: Medium (780 chars)

Pattern Match:
  "Surprising Fact" hook → above average
  "Tuesday" posting → confirms pattern
  "Personal Reference" → +40% (Low Confidence, n=4)

ICP Insight:
  Top Engagers: CTOs (30%), Developers (40%), PMs (20%)

Next snapshot in 5 days (Day 14).

Batch (/analyze)

1. All posts in data/posts/ with missing snapshots or without "Analyzed" status
2. For each: Run analysis
3. At the end: Pattern refresh across all archived posts

Pattern Refresh (/analyze patterns)

Only on archived posts (data/posts/archive/):

1. For each dimension: Calculate avg, sample size, success rate
2. Test combination patterns
3. Update pattern files in data/patterns/
4. Experiment evaluation (if experiments are running)

ICP Sharpening (/analyze icp)

1. Aggregate top demographics across all posts
2. Compare with configured ICP
3. Update ICP files in data/icp/
4. Show delta

Rules

• Respect lifecycle — Active posts with caveat, patterns only from archived
• 3 snapshots before status → "Analyzed"
• Median as baseline (more robust)
• Patterns only with data — min 3 posts per dimension
• Combos only with min 3 posts per combination
• Honest — if no patterns, say so